Machine Learning Portfolio

A collection of machine learning projects covering regression analysis, CNN image classification, RNN time-series forecasting, dimensionality reduction, and neural networks.

---

📁 Summative Projects

1. Boston House Price Prediction

📊 Regression Analysis | 🏠 Real Estate

Predicting median house prices in Boston using various regression techniques.

• Models: Simple Linear, Ridge, Lasso, Elastic Net Regression
• Key Features: RM, LSTAT, DIS, AGE, PTRATIO
• Best Model: Simple Linear Regression (R²: 0.80, RMSE: 2.25)
• Insights: Number of rooms (RM) strongly increases prices, while lower-status population (LSTAT) decreases them

Files:

• 💻 Code: Inventado_Lab1_Boston_House_Price_Prediction.ipynb
• 📄 Report: Inventado-Boston_House_Price_Report.pdf
• 📊 Datasets:
  • boston (1).csv (All columns)
  • boston_selected_features_withDIS.csv (Selected columns)

---

2. Hair Type Classification with CNN

🖼️ Image Classification | 💇 Hair Types

Multi-class classification of hair types (Curly, Straight, Wavy) using Convolutional Neural Networks.

• Models: Custom CNN architectures built from scratch
• Dataset: 985 images across 3 classes
• Best Model: 4 Conv layers, 128 neurons, learning rate 1e-4
• Results: 88% training accuracy, 61% validation accuracy, 0.81 ROC-AUC

Files:

• 💻 Code: Summative Lab Exercise 3 - Identifying Hair Types.ipynb
• 📄 Report: Inventado_Valles_Summative-Lab-Exercise_Identifying_Hair_Types.pdf
• 📊 Dataset: Hair Dataset

---

3. Stock Price Prediction with RNN

📈 Time Series Forecasting | 💹 Finance

Predicting Amazon and IBM stock prices using Recurrent Neural Networks.

• Models: LSTM, GRU, Bi-RNN architectures
• Data: 2006–2018 stock price history
• Best Model: GRU with 50/20 units, Adam optimizer, rolling averages
• Results: R² > 0.98, RMSE: 13.80 (AMZN), 1.68 (IBM)

Files:

• 💻 Code: Summative Lab Exercise 4_Predicting Stock Prices.ipynb
• 📄 Report: Inventado_Valles_Summative-Lab-Exercise-Predictin-Stock-Prices.pdf
• 📊 Datasets:
  • AMZN_2006-01-01_to_2018-01-01.csv
  • IBM_2006-01-01_to_2018-01-01.csv

---

🧪 Formative Lab Exercises

Formative Lab Exercise #2: Linear Regression & Logistic Regression

📉 Supervised Learning | 📊 Regression & Classification

• Linear Regression Task: Predicting student performance metrics.
• Logistic Regression Task: Diagnostic classification of breast cancer tissue samples.
• Files:
  • 💻 Code: Inventado_Formative Lab Exercise 2 HandsOn Linear Regression and Logistic Regression.ipynb
  • 📊 Linear Regression Dataset: CSE_student_performances.csv
  • 📊 Logistic Regression Dataset (cleaned): breast_cancer_cleaned.csv

---

Formative Lab Exercise #4: Multilayer Perceptron (MLP)

🧠 Deep Learning | 💰 Income Classification

• Description: Construction and training of a Multilayer Perceptron neural network to classify individual income levels based on census attributes.
• Files:
  • 💻 Code: Inventado - Formative 4 - Multilayer Perceptron (MLP).ipynb
  • 📊 Dataset: UCI Adult Income Dataset (adult.data)

---

Formative Lab Exercise #5: Principal Component Analysis (PCA)

📐 Unsupervised Learning | 🖼️ Dimensionality Reduction

• Description: Demonstrates Principal Component Analysis (PCA) for dimensionality reduction using the Fashion-MNIST dataset. The 28×28 pixel images are flattened into 784-dimensional feature vectors and normalized to zero mean and unit variance. PCA is applied across varying component sizes (n = 100, 50, 10) to reconstruct images, and Mean Squared Error (MSE) is evaluated to measure information loss.
• Files:
  • 💻 Code: Inventado - Formative 5 - Principal Component Analysis (PCA).ipynb
  • 📊 Dataset: Built-in tensorflow.keras.datasets.fashion_mnist

---

🛠️ Technologies Used

• Python with Jupyter Notebooks
• Machine Learning: Scikit-learn, TensorFlow/Keras
• Data Analysis: Pandas, NumPy, Matplotlib, Seaborn
• Specialized Models: CNNs, RNNs (LSTM/GRU), Multilayer Perceptron, PCA, Linear/Logistic Regression

---

📊 Key Skills Demonstrated

• Data preprocessing, normalization, and feature engineering
• Dimensionality reduction and variance analysis with PCA
• Hyperparameter tuning and deep learning optimization
• Supervised learning with regression analysis and binary/multiclass classification
• CNN architecture design for computer vision applications
• Time-series modeling using Recurrent Neural Networks (LSTM/GRU)
• Comprehensive evaluation metrics interpretation (R², RMSE, ROC-AUC, Reconstruction MSE)

---

👥 Authors

Charles Fredric G. Inventado & James Vincent V. Valles (For Lab 3 and Lab 4)
University of Santo Tomas - College of Information and Computing Sciences